The purpose of this report is to document the current form and methodologies used by the GoldFactory threat actor. The information documented is then used by Cyber Security Associates Ltd (CSA) Cyber Analysts to detect and hunt for the threat within the client environment through the use of our supported SIEM’s BorderPoint, Microsoft Sentinel and LogRhythm and advise on counter measures to monitor and detect for the subject threat.
This report documents the threat group GoldPickaxe and their TTPs (Tactics, Techniques and Procedures). Containing recommendations to help detect and mitigate the threat. The report also includes references where information within this report was identified from.
GoldFactory has created a highly advanced Trojan application that is designed to exfiltrate facial recognition data from a victims phone to an attacker operated database. This data is then used within an artificial intelligence workflow to create ‘deepfakes’ of victims and gain access to their facial recognition secured banking applications. This is the first recoded instance of this type of virus for iOS devices due to their solid utilisation of safety protocols and best practices. There are little ways to protect against it apart from maintaining awareness and not blindly trusting emails or text messages as convincing as they may be. Be on particular lookout for messages from commonly trusted entities such as banks or pension funds asking to verify documents or click and download from links.
Social Engineering
Social Engineering is a well-known tactic used by cyber threat actors to leverage peoples willingness to help or trust, people are often willing to assist or conform to strangers requests due to their own kindness or due to their perceived authority. For example, offering to lend someone trusted money or an account credentials because they are in a time of need ‘I am the prince of X, I have unfortunately lost the key to my safe, to get another one I need £100 but I will share my wealth with you, I promise!’ or the age old tale of ‘I am calling from Microsoft, we suspect you have a virus on your PC, please buy me a gift card so that I can remove it for you’. Luckily, most people can easily see that both of those examples are bad attempts at fraud. However, as with anything in technology there have been improvements to the efficiency and an added level of professionalism to these attempts. Specialist crime groups have been created that are dedicated to making these phishing attempts as good as possible, unfortunately the success rate has been increasing [8].
Phishing
Phishing is a type of social engineering that focuses on an attacker pretending to be a reputable entity; a member of IT Services asking you to click a suspicious link or asking for your password and login due to a system upgrade. These are just some examples of phishing attempts. Attackers will often use emails as an easy way to distribute phishing emails and use attachments or links to get a way in.
Smishing
Smishing is a type of phishing that focuses on deceiving targets via text messages to appear more personal than phishing emails, often spoofing phone numbers of banks or other reputable entities into the text field. Texting applications often lack the advanced spam detection capabilities emails have and are often an easier way of fooling targets into clicking links or even installing applications due to users placing more trust into this way of communication. A popular example of this in the UK is ‘you have missed a Royal Mail parcel delivery, please click this link to arrange a re-delivery’ [9].
Apple TestFlight Platform
Apples Test Flight platform is an easy way for developers to beta test their applications without having to go through Apples rigorous testing for them to be signed off and allowed onto the App store. This way developers can test their apps with a small group of chosen users which will test the application for them in a controlled manner, with the added benefit of being able to send the users a URL that will let them download the application. This ease of use for developers can easily be taken advantage of by malicious actors. Due to the lack of testing to applications on Apples TestFlight platform, it makes it significantly easier for a compromised application to make its way on there. From a phishing perspective, this makes it incredibly easy to infect a device with a genuine looking link and webpage- all without having to create any back end infrastructure to host the application and making a believable webpage.
Mobile Device Management
Mobile Device Management (MDM) is an Apple device management solution for maintenance and security of a fleet of devices that lets admins install, change and modify all aspects of a device such as application deployment or setting changes. Its Microsoft counterpart is known as Intune.
However, due to its potential it has also been utilised by malicious actors to install malware as uncovered by W. Mercer et al [2]. The authors discovered a malicious MDM platform that was loading fake applications onto smartphones. The attackers exploited a common MDM feature and used an iOS profile to hide and disable legitimate versions of apps, forcing users to interact with the malicious stand-ins that were disguised as applications such as ‘Safari, WhatsApp and Telegram’. The profile abused a section of MDM used to hide applications with an age restriction, by setting the age lower than the 12 and 17 required for WhatsApp and Telegram. The age of 9 was used in this scenario and due to this, the legitimate applications were restricted on the device and only their malicious counterparts remained accessible and visible to the users.
Rise of Online Banking and Law changes in Asia
Due to the global situation in 2020, online banking increased in popularity exponentially and due to its popularity it became a profitable target for cyber criminals. Due to growing security concerns Thai policy makers have required banks to enforce MFA via facial recognition if transfers over a certain amount are attempted.
The process of this operation is simple and very effective [Figure 1]
Figure 1: Biological MFA flowchart
Due to the maturity of facial recognition technology, this is a simple and effective solution that circumvents the common issues with passwords such as password sharing and setting weak passwords.
Tactics, Techniques, and Procedures (TTPs) describes the actions, behaviours, processes and strategies used by malicious adversaries that engage in cyber-attacks.
Tactics will outline the overall goals behind an attack, including the strategies that were followed by the attacker to implement the attack. For example, the goal may be to steal credentials. Understanding the Tactics of an adversary can help in predicting the upcoming attacks and detect those in early stages
Techniques will show the method that was used to engage in the attack, such as cross-site scripting (XSS), manipulation through social engineering and phishing, to name a few. Identifying the Techniques used during an attack can help discover an organisation’s blind spots and implement countermeasures in advance.
Procedures will describe the tools and methods used to produce a step-by-step description of the attack. Procedures can help to create a profile for a threat actor or threat group. The analysis of the procedures used by the adversary can help to understand what the adversary is looking for within their target’s infrastructure.
Analysts follow this methodology to analyse and define the TTPs to aid in counterintelligence. TTPs that are described within this research are based of the information which CSA analysts have been able to identify prior to the release of this document. The threat may change and adapt as it matures to increase its likelihood of evading defence.
GoldPickaxe is a sophisticated Trojan virus aimed at iOS devices running 17.4 or below, there are two ways in which it can infect the device, both of which include the user clicking a link, downloading and finally approving the installation. This happens either via an MDM profile or via a TestFlight URL. This is then used to install a legitimate looking application designed to fool the user into providing further information via the Trojan. The device is open to receiving commands via its Command-and-Control server. The information harvested is then used to create deep fake videos to pass MFA and log into banking accounts.
In this section the attack methodology will be discussed and laid out. This section assumes the user assists the attackers by successfully following prompts and clicking links on their compatible iPhone running iOS 17.4 or below. It also assumes the user has the password to the iCloud account associated to the device to enable the installation of the MDM profiles/applications depending on the attack methodology. This section is based on the findings of Group-IB [1] [Figure 2].
MITRE developed the Adversarial Tactics, Techniques and Common Knowledge framework (ATT&CK), which is used to track various techniques attackers use throughout the different stages of cyberattack to infiltrate a network and exfiltrate data.
The framework defines the following tactics that are used in a cyberattack:
After the initial development of rapport with the victim the attacker will attempt to compromise the user device. There are two possible ways of infection via GoldPickaxe.iOS; either by the user being lured to install an application via TestFlight or following a malicious URL to another webpage controlled by the attacker which will download and attempt to enable an MDM profile on the victims device.
These are both examples of techniques T1565 (Phishing) and T1119 (Trusted Relationship).
If the user installs the application via TestFlight, the user will follow the testflight.apple.com/join/ URL and download the trojan as well as the genuine application onto the device. This is now a compromised target and will follow onto Phase 2.
If the user installs the MDM profile, the user follows the URL link sent to them, the MDM profile is automatically downloaded and the user is asked for permission to install it. After this is successful, the device will download the malicious application via Safari browser and install GoldPickaxe.iOS silently on the device. This is now a compromised target and will follow onto Phase 2.
Both of the techniques outlined utilise the T1204 (User execution) approach from the attack matrix as they rely on the user to execute the packages.
At this stage the threat actor has full and unrestricted access to the device, it does however require user interaction within the application to create the data the attacker is after. These actions will be paired with prompts by the attacker via whatever way the initial point of contact was, for this we assume it was by text.
The attacker will message the user to open the application and provide verification within it, this can be done by; recoding a short video, requesting photos of ID’s or other documents. The application also has further abilities such as interception of text messages and web traffic to and from the device.
At this stage the attackers will perform multiple examples of collection techniques, mainly: T1113 (Screen capture), T1115 (Clipboard Data), T1005 (Data from Local System) and they will finally utilise T1560 (Achieve Collected Data) for ease of exfiltration. The data created by the user will be downloaded onto the device for a later exfiltration stage which is detailed in Phase 3.
Within this stage the data that was harvested from the compromised individual is sent back to the attackers controlled database. This type of communication is controlled by sending the correct commands to the device via its WebSocket located at 8383. The data sent back regarding the specific command will be transmitted via HTTP API. This is an example of the command and control technique T1071 (Application Layer Protocol) due to the usage of normal protocols and the usage of T1090 (Proxy). However, there is also another communication channel specifically designed for exfiltration of data into a cloud bucket storage location. This being an example of T1537 (Transfer Data to Cloud Account).
This is one of the few indicators of compromise (IOC’s) for this trojan application as communication with specific URL’s can be used as a confirmation of a devices compromise status. The commands sent to the devices as well as hash values and URL’s accessed are included within the Indicators of compromise section.
The information sent back to the attacker can include items from the users gallery, SMS messages, captures of the users face and network activity logs. This will be used in the final phase of the attack, Phase 4.
The final stage is where the data manipulation and the utilisation occur. It is believed by Group-IB that the attackers utilise the identification documentation as well as the recorded short video as sources for deep fake creation purposes. Due to the creation process, the more source files and angles of a person you have the more genuine the deep fake video will be. The final source files will be layered over an attackers face which will match up with the prompts used by banking apps in order to pass verification as the victim.
There are a multitude of options for deep fake creation [4] ranging from reface [5] which is an online platform to standalone applications such as DeepFaceLAB 2.0 [6] which can utilise Nvidia and AMD graphics cards to further enhance the work flow and level of realism of the final work. The standalone option also has the added benefits of being able to use advanced shaders and other addons to create hyper realistic deep fake videos.
At this step the attackers have successfully compromised the account and can now exfiltrate the funds or apply for finance. The attackers are suspected to use other devices that proxy into the victims network to circumvent regional checks from banking applications which is an example of T1090 (Proxy).
The cyber kill chain is a process that traces the stages of a cyberattack. This starts at the early reconnaissance stages that eventually leads to data exfiltration.
The kill chain can help one to understand and combat ransomware, advanced persistent threats (APTs) and security breaches.
The cyber kill-chain defines the following tactics that are
In conclusion, due to the significant capabilities of the Trojan application and it being the first of its kind for iOS devices, it would be foolish to assume that it will not be shared between threat groups. This means that in the near future more countries will be targeted with advanced phishing campaigns looking to take advantage of users. As the malicious MDM profile approach is very powerful and essentially a ‘golden’ ticket for attackers, it requires a certain amount of vigilance from users.
However, due to it requiring the assistance of the devices owner in providing sensitive information and pictures/ videos, it is unlikely that many people will fall for it or even have the data on their device in the first place.
In the coming months and years, we are likely to see more Trojans being developed for the iOS ecosystem due to its prolific use. It is also likely that these iterations will build on previous versions of the Trojan. This means we will see an increase in capabilities and potentially even more advanced installation procedures like silent installation etc without the need for the users assistance.
Due to the attack vectors used by the Trojan application there are only a few things that need to be done to stay protected and secure. The best defence against any Trojan application is to always download applications from secure sources and to always remain suspicious of any communications before validating their sources, this includes applying MDM profiles to devices from anyone other than a known system admin. Obviously as the validation process becomes increasingly more difficult, it is advisable to use multiple sources to confirm. An example of this would be physically going into a branch of your bank and verifying if they really need more documentation or even calling them to do the same.
Due to the significant capabilities of the trojan package it is likely that infected users are only able to verify their status via their Antivirus software matching IOC’s for GoldPickaxe. Performing regular antivirus scans of devices ensures that any downloads are scanned for malicious payloads in real time to prevent further instances of malware.
If a device has been deemed as infected it is best to factory reset it to make sure any leftover files are destroyed. It is also recommended to change all passwords on all accounts that were signed into on the device as their status may have been compromised.
As a final best practice it’s advisable to regularly check for software updates as they often include patches and security updates which help to keep devices safe and optimised.
Indicators of Compromise GoldPickaxe iOS Trojan [Table 1] [11].
Based against Mitre ATT&CK Framework [12] [Table 2]
[1] Face Off: Group-IB identifies first iOS trojan stealing facial recognition data
[2] Malicious MDM: Let's Hide This App
[3] Let's dig deeper: dissecting the new Android Trojan GoldDigger with Group-IB Fraud Matrix
[5] Realistic face swaps in a click
[6] DeepFaceLab 2.0 Installation Tutorial
[7] Google Play malware clocks up more than 600 million downloads in 2023
[8] The Latest 2024 Phishing Statistics (updated May 2024)
[9] Typical online scams to look out for
[10] Lost 3 billion because of facial recognition camera scanning on fake public service software
[11] ACTIVIDAD MALICIOSA | Relacionada con GoldPickaxe 15-02-2024
[12] ATT&CK | Matrix for Enterprise